Advertisement


Compu-Picks 2012 Analysis: Week Two Upsets

Mr Pac Ten
Posted Sep 11, 2012


2012 Compu-Picks Analysis: Looking at the impact of key week two upsets

One new piece of analysis I'm doing this year is looking at some of key matchups and what Compu-Picks thinks the results of these matchups will say about the rest of the season. I've done this a couple times already (the latest was a look at key week two matchups), and this a continuation of the series.

Now that week two is in the books, I'm going to take a look at some key upsets and what we can learn from them. Please note, however, that this analysis is entirely based on the preseason simulations; the actual scores, stats, levels of dominance etc. have NOT been considered for this analysis.

I'll start with the most extreme upset of the weekend, UL-Monroe beating Arkansas as about a 30-point underdog. The below table shows the win likelihoods for each game on ULM's schedule, as well as an aggregate expected win value (one for all games, one for all games excluding Arkansas; this is meant to highlight the predictive impact that a win here has on all other games).

In addition, there are three columns here: the right-most column shows the original predictions, knowing nothing about any actual season results. The other two columns show projections given either a win or a loss against Arkansas. As you can see, beating Arkansas is worth nearly two full wins for ULM on top of the actual game result, for a total impact of almost three. That's how much we can potentially learn from an upset of this magnitude against ULM's schedule.

This is especially potent since it's ULM's first result; had ULM played a few other games before this one, we'd have more data on them and the impact of this specific game would be less. For instance, if they'd played Auburn, Baylor, Tulane and MTSU and gone 2-2, then even given a win here it would be fairly clear that there's no chance they're secretly an elite team, and just as clear that they're not atrocious. Without any games played, though, both possibilities are more on the table, and therefore this data point means more (at least, until they get around to playing a couple more games and we start to see more data).

La.-Monroe 2012 Season Projections based on W/L vs Arkansas
Arkansas Result W L Any
Win % vs
Arkansas 100% 0% 7%
Auburn 8% 2% 3%
Baylor 50% 27% 29%
Tulane 69% 50% 51%
M. Tenn. St. 81% 70% 71%
Fla. Atlantic 91% 77% 78%
W. Kentucky 44% 22% 24%
South Alabama 95% 88% 88%
La.-Lafayette 65% 45% 47%
Arkansas St. 52% 32% 33%
North Texas 85% 68% 69%
FIU 54% 28% 30%
Total E(Wins) - except for Arkansas 6.93 5.10 5.23
Total E(Wins) - including Arkansas 7.93 5.10 5.30

It's a similar story for Arkansas, except in reverse. Here, the Hogs have already played a game against a AA team, so instead of three columns we see four. Again, the right-most column is the season projection without any actual data. The next column is the adjusted projection given a win over their AA opponent. Given that the Hogs were basically given 100% odds of winning that game (the actual projected upset chance was 0.4%, which rounded down to 0%), it's not suprising that there was basically no movement in their win projections given the AA win.

There is, however, substantial movement given the loss to UL-Monroe. As you can see, there is nearly a two-game difference in Arkansas's win projections given a win versus a loss here (note that both the W and L columns are given the AA win; I've already thrown out the handful of scenarios where the Hogs lost the AA game, since we already know that didn't happen).

Losing to ULM means that a true disaster season is absolutely on the table, and a surprising BCS run simply isn't going to happen. If they're not good enough to beat ULM, they're not likely to be good enough to beat Bama, South Carolina, Miss St, or LSU, much less all of them. Personally, I think that a win expectation of 4.34 is too low, but it's still a good indication of just how much a loss like this can tell you about a team's chances going forward.

Arkansas 2012 Season Projections based on W/L vs La.-Monroe given 1-0 start
La.-Monroe Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 100%
Win % vs
La.-Monroe 100% 0% 93% 93%
Alabama 21% 9% 20% 20%
Rutgers 68% 45% 67% 66%
Texas A&M 54% 30% 52% 52%
Auburn 23% 7% 22% 22%
Kentucky 89% 73% 88% 88%
Mississippi 88% 71% 87% 87%
Tulsa 77% 55% 76% 76%
South Carolina 33% 13% 31% 31%
Mississippi St. 33% 14% 31% 31%
LSU 38% 16% 36% 36%
Total E(Wins) - except for La.-Monroe 6.24 4.34 6.11 6.10
Total E(Wins) - including La.-Monroe 7.24 4.34 7.04 7.02

Next up is Utah-Utah St. Again, this one had a major impact. While the line was only -7, Compu-Picks gave Utah St a much smaller shot at victory. So when they pulled the upset, there was a more substantial revision to the projections for both teams than a simple 7-point upset would otherwise suggest.

Utah 2012 Season Projections based on W/L vs Utah St. given 1-0 start
Utah St. Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 100%
Win % vs
Utah St. 100% 0% 90% 90%
BYU 57% 35% 55% 55%
Arizona St. 92% 80% 91% 90%
USC 56% 34% 54% 54%
UCLA 59% 39% 57% 57%
Oregon St. 66% 42% 63% 63%
California 92% 77% 90% 90%
Washington St. 93% 85% 92% 92%
Washington 56% 32% 54% 54%
Arizona 96% 87% 95% 95%
Colorado 96% 88% 95% 95%
Total E(Wins) - except for Utah St. 8.64 7.00 8.47 8.45
Total E(Wins) - including Utah St. 9.64 7.00 9.36 9.35

Utah 2012 Season Projections based on W/L vs Utah St. given 1-0 start
Utah St. Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 100%
Win % vs
Utah St. 100% 0% 90% 90%
BYU 57% 35% 55% 55%
Arizona St. 92% 80% 91% 90%
USC 56% 34% 54% 54%
UCLA 59% 39% 57% 57%
Oregon St. 66% 42% 63% 63%
California 92% 77% 90% 90%
Washington St. 93% 85% 92% 92%
Washington 56% 32% 54% 54%
Arizona 96% 87% 95% 95%
Colorado 96% 88% 95% 95%
Total E(Wins) - except for Utah St. 8.64 7.00 8.47 8.45
Total E(Wins) - including Utah St. 9.64 7.00 9.36 9.35

Utah St. 2012 Season Projections based on W/L vs Utah given 1-0 start
Utah Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 90%
Win % vs
Utah 100% 0% 11% 10%
Wisconsin 25% 12% 14% 13%
Colorado St. 81% 62% 64% 61%
UNLV 96% 90% 90% 88%
BYU 15% 5% 6% 6%
San Jose St. 77% 62% 64% 61%
New Mexico St. 99% 96% 96% 95%
Texas-San Antonio 92% 84% 85% 82%
Tex St-San Mar 94% 90% 90% 88%
Louisiana Tech 37% 21% 23% 21%
Idaho 94% 86% 87% 85%
Total E(Wins) - except for Utah 8.10 7.07 7.19 6.89
Total E(Wins) - including Utah 9.10 7.07 7.30 6.99

Maryland-Temple is another game where a meaningful upset had a material effect on the win projections:

Maryland 2012 Season Projections based on W/L vs Temple given 1-0 start
Temple Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 86%
Win % vs
Temple 100% 0% 24% 22%
Connecticut 58% 42% 46% 43%
West Virginia 10% 3% 4% 4%
Wake Forest 61% 46% 50% 46%
Virginia 52% 36% 40% 36%
N. Carolina St. 51% 37% 40% 37%
Boston College 13% 6% 8% 7%
Georgia Tech 29% 15% 19% 17%
Clemson 8% 3% 4% 3%
Florida St. 7% 2% 3% 2%
North Carolina 55% 37% 42% 38%
Total E(Wins) - except for Temple 4.44 3.26 3.54 3.20
Total E(Wins) - including Temple 5.44 3.26 3.78 3.42

Temple 2012 Season Projections based on W/L vs Maryland given 1-0 start
Maryland Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 97%
Win % vs
Maryland 100% 0% 79% 78%
Penn St. 62% 45% 59% 58%
South Florida 50% 31% 46% 45%
Connecticut 61% 45% 58% 57%
Rutgers 39% 26% 36% 35%
Pittsburgh 39% 23% 35% 35%
Louisville 31% 18% 29% 28%
Cincinnati 51% 33% 48% 47%
Army 60% 44% 56% 56%
Syracuse 78% 64% 75% 74%
Total E(Wins) - except for Maryland 5.71 4.29 5.42 5.31
Total E(Wins) - including Maryland 6.71 4.29 6.21 6.10

Next up is Rice-Kansas. Here, the upset was pretty meaningful, but since most games for both teams are either very likely wins or very likely losses, the impact was actually not that large:

Rice 2012 Season Projections based on W/L vs Kansas given 1-0 start
Kansas Result W Given 0-1 L Given 0-1 Any Given 1-0 Any
UCLA - already played 0% 0% 0% 15%
Win % vs
Kansas 100% 0% 28% 31%
Louisiana Tech 22% 13% 16% 18%
Marshall 63% 48% 53% 55%
Houston 48% 34% 38% 41%
Memphis 86% 75% 78% 80%
Texas-San Antonio 92% 84% 87% 88%
Tulsa 16% 11% 12% 14%
Southern Miss 35% 24% 27% 30%
Tulane 57% 42% 46% 49%
SMU 55% 41% 45% 48%
UTEP 54% 39% 43% 46%
Total E(Wins) - except for Kansas 5.28 4.11 4.44 4.83
Total E(Wins) - including Kansas 6.28 4.11 4.73 5.14

Kansas 2012 Season Projections based on W/L vs Rice given 1-0 start
Rice Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 92%
Win % vs
Rice 100% 0% 71% 69%
TCU 8% 4% 7% 7%
Northern Illinois 31% 15% 26% 25%
Kansas St. 18% 11% 16% 15%
Oklahoma St. 19% 10% 17% 16%
Oklahoma 1% 0% 1% 1%
Texas 5% 1% 4% 3%
Baylor 30% 18% 26% 25%
Texas Tech 26% 12% 22% 20%
Iowa St. 54% 39% 50% 47%
West Virginia 6% 1% 4% 4%
Total E(Wins) - except for Rice 2.99 2.11 2.74 2.56
Total E(Wins) - including Rice 3.99 2.11 3.45 3.25

Next up is UNC-Wake, a game where the W/L result went as predicted (Compu-Picks took Wake), but not by nearly as much as predicted (Compu-Picks thought Wake would win easily). Here's a good example of when to take the system's conclusions with a grain of salt; since the result was very nearly a loss, the "true" interpretation of the win likelihoods going forward is probably a midpoint between what you'd expect given a win (regardless of score) and a loss (again regardless of score).

North Carolina 2012 Season Projections based on W/L vs Wake Forest given 1-0 start
Wake Forest Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 89%
Win % vs
Wake Forest 100% 0% 34% 32%
Louisville 22% 10% 14% 13%
East Carolina 66% 50% 56% 53%
Idaho 89% 84% 86% 83%
Virginia Tech 27% 17% 20% 18%
Miami (Fla.) 31% 19% 23% 22%
Duke 35% 22% 27% 25%
N. Carolina St. 52% 36% 41% 39%
Georgia Tech 28% 18% 21% 20%
Virginia 53% 37% 43% 40%
Maryland 74% 60% 65% 62%
Total E(Wins) - except for Wake Forest 5.78 4.53 4.95 4.63
Total E(Wins) - including Wake Forest 6.78 4.53 5.30 4.95

Wake Forest 2012 Season Projections based on W/L vs North Carolina given 1-0 start
North Carolina Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 94%
Win % vs
North Carolina 100% 0% 70% 68%
Florida St. 3% 1% 2% 2%
Army 65% 50% 61% 59%
Duke 54% 37% 49% 47%
Maryland 60% 45% 56% 54%
Virginia 59% 44% 54% 53%
Clemson 17% 9% 14% 14%
Boston College 30% 17% 26% 25%
N. Carolina St. 43% 27% 38% 37%
Notre Dame 17% 7% 14% 13%
Vanderbilt 26% 15% 23% 22%
Total E(Wins) - except for North Carolina 4.74 3.52 4.37 4.21
Total E(Wins) - including North Carolina 5.74 3.52 5.07 4.89

Finally we come to the last upset of Saturday night: Arizona over Oklahoma St. Here, Compu-Picks was simply wrong, and as such, you can see a pretty big shift in the projections for each school. It's not yet clear whether the story here is that Arizona is surprisingly good, Oklahoma St is surprisingly bad, it was just a wacky upset, or some combination of the three, but even so you can see a pretty big shift as a result of the upset.

Oklahoma St. 2012 Season Projections based on W/L vs Arizona given 1-0 start
Arizona Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
AA - already played 100% 100% 100% 100%
Win % vs
Arizona 100% 0% 89% 89%
La.-Lafayette 92% 80% 91% 91%
Texas 27% 14% 25% 25%
Kansas 86% 68% 84% 84%
Iowa St. 92% 78% 91% 90%
TCU 43% 25% 41% 41%
Kansas St. 61% 37% 58% 58%
West Virginia 58% 35% 55% 55%
Texas Tech 82% 67% 80% 80%
Oklahoma 15% 6% 14% 14%
Baylor 79% 51% 76% 76%
Total E(Wins) - except for Arizona 7.35 5.61 7.16 7.14
Total E(Wins) - including Arizona 8.35 5.61 8.05 8.03

Arizona 2012 Season Projections based on W/L vs Oklahoma St. given 1-0 start
Oklahoma St. Result W Given 1-0 L Given 1-0 Any Given 1-0 Any
Toledo - already played 100% 100% 100% 49%
Win % vs
Oklahoma St. 100% 0% 16% 11%
Oregon 2% 0% 1% 0%
Oregon St. 52% 32% 35% 27%
Stanford 9% 3% 4% 3%
Washington 40% 23% 26% 20%
USC 29% 12% 15% 10%
UCLA 21% 10% 12% 8%
Colorado 93% 82% 84% 77%
Utah 18% 6% 8% 5%
Arizona St. 82% 70% 72% 64%
AA 97% 94% 95% 90%
Total E(Wins) - except for Oklahoma St. 5.43 4.32 4.50 3.53
Total E(Wins) - including Oklahoma St. 6.43 4.32 4.67 3.64

I thought about also showing this week's AA losses, but there were so many that it would have just clogged the article.

There are a few important notes and caveats I need to make about this model:

1) Compu-Picks does not endorse implicitly or explicitly any form of illegal gambling. Compu-Picks is intended to be used for entertainment purposes only.

2) No guarantee or warranty is offered or implied by Compu-Picks for any information provided and/or predictions made.

3) This preseason model is primarily based on the main compu-picks model. Essentially, it attempts to predict how well a team will rate given its rating history, as well as a number of other data points, such as returning starters, draft talent lost, turnovers, recruiting, etc. This means, among other things, that the rankings are power rankings based on how good a team projects to be, as opposed to a more cynical (though accurate) model that attempts to project how the BCS will rank a team by making adjustments to favor those with easy schedules and punish those with tough schedules.

2012 Compu-Picks Blog

Questions, comments or suggestions? Email me at cfn_ms@hotmail.com

Follow cfn_ms on Twitter

Related Stories
Compu-Picks 2012 Analysis: Week Two
 -by CollegeFootballNews.com  Sep 10, 2012
Todd Monken: Lunt Has Been Impressive
 -by GoPokes.com  Sep 10, 2012
Bill Young Anxious To Get Back On Field
 -by GoPokes.com  Sep 10, 2012
























Unauthorized use of ad tag